special issue: ica commission on mountain cartography edited … · 2015. 9. 14. · ing engine...

CARTOGRAPHICA Volume 38 / Numbers 1&2 / Spring/Summer 2001

Endorsed by theinternational cartographic association /association cartographique internationale

and endorsed as the journal of thec a n a d i a n c a r t o g r a p h i c a s s o c i a t i o n / a s s o c i a t i o n c a n a d i e n n e d e c a r t o g r a p h i e

and supported by grants provided by thes o c i a l s c i e n c e s a n d h u m a n i t i e s r e s e a r c h c o u n c i l

editor 1965–1994 and founder, bernard v gutsell

Editor : Brian Klinkenbergu n i v e r s i t y o f b r i t i s h c o l u m b i a , v a n c o u v e r , b r i t i s h c o l u m b i a

© University of Toronto Press Incorporated 2003issn 0317-7173

Printed in Canada

i n t e r n a t i o n a l p u b l i c a t i o n s o n c a r t o g r a p h y

Special Issue: ICA Commission on Mountain CartographyEdited by Lorenz Hurni, Karel Kriz, Tom Patterson, and Roger Wheate

89

CARTOGRAPHICA, VOLUME 38, # 1&2, SPRING/SUMMER 2001

DEM Manipulation and 3-D Terrain Visualization:Techniques Used by the U.S. National Park Service

tom patterson

U.S. National Park Service / Harpers Ferry Center / Harpers Ferry / WV / USA

AbstractManipulating digital elevation model (dem) surfaces, likepliable modelling clay, enhances the appearance and leg-ibility of 3-D topography on maps. The U.S. National ParkService (nps) uses the familiar image-editing tools inAdobe Photoshop to manipulate raster dem data. Export-ing modified dem data with the help of freeware andshareware utilities allows subsequent rendering of final3-D scenes in Corel Bryce. Techniques to be discussed in-clude topographic substitution – a method for reverse en-gineering present-day landscapes into the past orprojecting them into the future; selective vertical exag-geration; resolution bumping – a technique developedspecifically for improving the legibility of high-mountainlandscapes; painting and filtering effects; and, borrowingfrom the traditional masters of landform depiction, cre-ating 3-D scenes that emulate the panoramas of HeinrichBerann and the spherical over-the-horizon views of Rich-ard Edes Harrison by warping the projection plane ofdems. The unique challenges of 3-D mountain mappingand the continuing pursuit of design excellence – a cor-nerstone of the nps cartographic program – are overarch-ing themes.

Introductionhis paper discusses the graphical techniques usedby the U.S. National Park Service (nps) in ma-nipulating digital elevation models (dems) for

the design and production of 3-D landscape visualiza-tions. After discussing approaches to dem manipulation(geographic information system versus graphical), anoverview of raster-based dem manipulation in AdobePhotoshop – my software of preference – is given. Nextand most important, a discussion of specific techniquesemphasizes how to produce 3-D landscape visualizationswith modified dems. The techniques are not step-by-steptutorials. Considering the frenetic pace of software revi-sions, tutorials would soon be out of date (Photoshop

T

was updated twice since this article was first written). In-stead the goal is to acquaint cartographers with just a fewof the many beneficial manipulations that can be ap-plied to dems, and encourage development of new tech-niques that will enable mountain landscapes to bevisualized in unimagined ways.

Over the past five years, the nps has increasingly re-lied on 3-D visualizations – static images, animations, andinteractive scenes – to portray the natural and culturalresources of the 387 units in the nps. Perhaps becausethey look more realistic, it is generally assumed by npsstaff that 3-D visualizations are more easily understoodby casual visitors than abstract two-dimensional (2-D)maps and illustrations. The types of 3-D visualizationscreated by the nps include geologic block diagrams, nat-ural science illustrations, hiking maps, mountain pano-ramas, and bird’s-eye views of cultural sites showingbuildings, landscaping, and vegetation set on topograph-ic models.

DEMs – the term is used generically here to describeall varieties of spatially arranged elevation data, regard-less of format – comprise the digital foundation for all3-D landscape visualizations. Despite the importance ofdems, cartographers and gis specialists generally hesitateto manipulate dems to enhance 3-D visualizations – incontrast to their willingness to modify vector data rou-tinely to enhance 2-D maps. Manipulations of dems areusually made only to edit imperfect data, such as system-atic banding and discrepancies in matching edges. Thishands-off approach to dems may have several causes: thelingering belief that dems, like the landscapes they repre-sent, are largely immutable; well-intentioned respect formaintaining the integrity of data created by others; andunfamiliarity with the necessary software applicationsand techniques, and the concept of dem manipulation it-self.

Why Manipulate DEMs?Practical reasons abound for manipulating dems beyondroutine editing of cosmetic data imperfections. As in allcartographic presentation, geographic reality and graph-ic reality are often at cross purposes. Just as vector datamust be simplified to be legible on a 2-D map, dems must

Tom Patterson, U.S. National Park Service, Harpers Ferry Center,Harpers Ferry, WV, USA, 25425-0050. Tel.: +1 304-535-6020. E-mail:[email protected]

90 tom patterson


be generalized at reduced scales to avoid creating topog-raphy that looks more like a noisy graphical texture thana 3-D landscape. This problem can be ameliorated withresolution bumping – a dem manipulation techniquethat yields legible small-scale 3-D topography in ruggedhigh mountains, such as the Alps, without sacrificing de-tail (Figure 5). On the other extreme, small topographicfeatures whose importance is disproportionate to theirsize can be given greater emphasis by using selective ver-tical exaggeration. For example, Puu Oo volcano on theisland of Hawaii has a surface area of only 10 hectaresbut has sheathed 105 square kilometres of Hawaii withlava since it began continually erupting in 1983. With se-lective vertical exaggeration, Puu Oo could be exaggerat-ed in height and made more discernable next to itsdominant neighbour, Mauna Loa, whose mass is greaterthan any other solitary mountain in the world.

The most remarkable dem manipulation techniquesmay be those that depict geologic processes. Startingwith a dem of an existing landscape, the user can eitherreverse-engineer topography to portray an earlier stageof landscape development or project it into the future.Like virtual modelling clay, dems are malleable; they al-low users to create an almost unlimited variety of deriva-tive or even entirely new landscapes. The challengewhen altering the topographic morphology of dems is todo so with the control and precision required for geo-graphic visualization.

The main objective of dem manipulation is not themanipulation of dems themselves, but the enhancementof 3-D visualizations generated from the altered data.DEM manipulation is simply another technique – notunlike linework generalization on vector maps, whichcartographers may use for presenting complex spatial in-formation to audiences in a more comprehensible form.DEM manipulation is not for everyone. It is a highly spe-cialized endeavour that requires users to possess a deepunderstanding of landscapes, 3-D visualization, and ex-pertise with the requisite software. Design sense and re-straint are critically important.

Approaches to DEM ManipulationThe software used to manipulate DEMs can generally becategorized as either gis or graphic. Although the appli-cations in each category have similarities (such as theirunderlying algorithms), they are used by entirely differ-ent professional cultures to produce different products.GIS applications, as most readers of this text are wellaware, appropriately emphasize spatial analysis and accu-racy, but the relatively staid 3-D graphical capabilities ofthese applications provide few tools for creatively manip-ulating dems. By contrast, graphical 3-D software applica-tions generally contain a plethora of tools, some withreal-time interactivity, for manipulating elevation data tocreate eye-catching special effects. The landscapes yield-ed by graphical software are used mostly for entertain-ment (computer games and movie special effects), fine

art, and fantasy/science fiction pursuits. Despite theseimaginative ends, graphical 3-D software tends to havelimited dem import capabilities. Imaginary fractally gen-erated terrains, some stunningly realistic, are the land-scape type of choice (Musgrave 1998). Geographicaccuracy, if addressed at all, is often an inconvenient af-terthought.

The nps approach to dems manipulation bridges thegap between the cartographic/gis and graphic 3-D cul-tures. Three-dimensional visualizations are created withCorel Bryce, a consumer-oriented, landscape renderingapplication most famous for its unconventional KaiKrause–inspired graphical user interface (http://www.corel.com). Because Bryce does not import geo-coded da-ta, helper utilities and a workaround procedure are usedto bring large-format dems into the program. However,once the dem is imported, a powerful suite of 3-D specialeffects exists; the interface is easy to use, and the render-ing engine yields exceptionally high-quality ray-tracedoutput in the form of raster images, animations, or Quick-Time Virtual Reality scenes. Although Bryce contains afascinating terrain editor, which allows dems to be manip-ulated interactively, its cramped interface and inability toedit precisely defined geographic areas limits effective-ness for making professional landscape visualizations(Webster 2001).

Manipulating DEMs in PhotoshopTo overcome the shortcomings of Bryce’s terrain editor,the nps uses Adobe Photoshop (http://www.adobe.com),the popular image-editing application, to manipulatedems before importing them into Bryce. Because the npsalready uses Photoshop as a go-between application forimporting large-format dems into Bryce, making dem ma-nipulations in Photoshop makes practical sense. Depend-ing on their format, in some cases Photoshop can opendems directly as raw binary files (row and column di-mensions must be known), or more easily with the help ofintermediary freeware utilities such as MacDEM (http://www.treeswallow.com) or MicroDEM (http://www.usna.edu/Users/oceano/pguth/website/microdem.htm), onMacintosh and PC respectively. When opened in Photo-shop, dems appear as 16-bit greyscale raster images con-taining 65,536 levels of height information. With 16-bitvertical resolution a user can depict topographic surfaceswith near-perfect smoothness – in contrast with the 256elevation levels available in 8-bit dems, which often yieldtopography with stair-stepped surfaces (Figure 1). To con-vert 16-bit greyscale dems from Photoshop format (.psd)to quadratic-sized files in portable greyscale map format(.pgm), which Bryce can then import, requires BSmooth(http://www.bsmooth.de), a Macintosh shareware utility,or pbm+, a Macintosh freeware Photoshop plug-in(http://www.nacis.org/cp/cp28/resources.html).

A dem opened in Photoshop appears as a ghostedgreyscale image, similar to an X-ray, with light-colouredpixels representing high elevations and dark pixels low-

dem manipulation and 3-d terrain visualization 91


lands. Although a greyscale dem in Photoshop looksnothing like the 3-D surface it will eventually becomewhen rendered in Bryce (Figure 2), being able to see itin pictorial form at all is a huge benefit when makingmanipulations. Changes actually can be seen, in contrastto the original dem – a text file of numbers that defiesvisualization.

Cartographers can use Photoshop’s image-editingand selection tools, more or less, to edit a greyscale demjust like any other image. Basic image transformations –cropping, downsampling (decreasing the resolution),rotating, and scaling – behave in the usual manner. Ap-plying the Gaussian blur filter to a dem results in asmooth generalized surface, while the noise filter addsroughness, and lightening or darkening the image ei-ther raises or lowers topography respectively – no sur-prises there. However, because they are data files,greyscale dems sometimes respond unexpectedly or det-rimentally to Photoshop manipulations that we are ac-customed to using with typical 2-D images. For example,the median filter, enormously useful for generalizing 2-Dshaded relief (Patterson 2001a), when applied to a demyields terrain with a creased surface that looks like crin-kled paper. The unsharp mask filter, another stalwart of2-D shaded-relief editing, produces unsightly needle-likeartifacts along prominent ridgelines and drainages.

There are notable disadvantages to manipulatingdems in Photoshop. Previewing an edited dem in 3-D in-volves exporting the dem from Photoshop, importing itinto Bryce, and then rendering the scene – a time-con-suming process when making repeated subtle edits.More problematic is Photoshop’s limited functionalitywhen using 16-bit images as opposed to “normal” 8-bitimages. Most items in the toolbar are disabled, as aremost filters and the layers palette – 16-bit functionality isakin to going back to Photoshop version 1.0.

Despite Photoshop’s diminished 16-bit functionality,it nevertheless contains all the tools needed to manipu-late dems with the accuracy required for cartographicvisualization. Clever workarounds, however, are oftennecessary to perform routine functions. For example,hard-edged and feathered selections can be used on 16-bit dems despite the fact that the magic wand and colorrange selection tools are disabled in 16-bit mode. Using

either of these selection tools requires duplicating a 16-bit dem file and saving it as an 8-bit proxy to be used onlyfor making selections. The selection boundaries canthen be dragged and dropped from the 8-bit file to the16-bit file. (Hint: temporarily inverting an interior selec-tion allows it to be precisely registered to a corner whendragged into the 16-bit image.) Cartographers can usesimilar techniques to import selections from rasterizedvector geo-data.

Photoshop’s painting tools are also disabled in 16-bitmode. However, the rubber stamp tool, which is ena-bled, is an effective substitute for the airbrush tool forpainting on dem surfaces. Data can be sampled andcloned between two open 16-bit Photoshop files. By fill-ing a proxy 16-bit file entirely with black or white andthen using the rubber stamp with a soft brush and set-ting the blending opacity to a low value, black or whitetones can be subtly transferred from the proxy imageonto the dem, lowering or raising elevations respectively.

Photoshop can be used in innumerable ways to ma-nipulate dems to enhance 3-D landscape visualization.The following examples emphasize creating static imag-es of 3-D landscapes to show some practical techniquesused by the nps. Although 3-D illustrations of terrain areused to depict the results of the techniques, readersneed to remember that the techniques are exclusivelyabout the manipulation of raw dem data (in greyscaleformat) rather than post-rendering touch-ups to illustra-tions.

GeneralizationDownsampling means decreasing the resolution or sizevia the Image Size dialog. Downsampling greyscale demsin Photoshop results in more generalized landscape sur-faces. For example, a dem downsampled from 1024 �1024 pixels to 512 � 512 pixels will have a quarter asmuch detail as the original and a smoother appearance.In the Preferences dialog, Photoshop offers three inter-polation methods for calculating pixel values whendownsampling or otherwise transforming images. Ingeneral, bicubic interpolation (the default) works bestfor straight downsampling. If the dem were also to be ro-

Figure 1. 3-D views of Mt. Rainier’s summit generated from 8-bit (left) and 16-bit (right) DEM data. Figure 2. 3-D geometric objects (top row) look considerably

different from the greyscale “DEMs” in Photoshop format (bottomrow) from which they were extruded.

92 tom patterson


tated, however, Nearest Neighbor interpolation wouldbe a better choice, because this faster interpolation es-chews anti-aliasing, thus preserving crisp-edged pixels atthe margins of the dem. Regardless of which interpola-tion method is used, upsampling – the procedure oppo-site to downsampling – which adds pixels to an image,should generally be avoided. Upsampling only increasesthe file size without increasing discernable detail.

Alternatively, applying the Gaussian blur filter to adem, a method that does not decrease the file size,achieves extremely smooth generalization. The smooth-ing effects of Gaussian blur filtering are different fromthose of downsampling; experimentation is required toachieve comparable levels of generalization between thetwo techniques.

Although generalization is most often applied global-ly throughout a dem, it can also be applied in graduatedamounts to achieve subtle visual effects. For example, in-creasing generalization from foreground to backgroundon a dem creates the optical illusion of depth when it isviewed in 3-D (Figure 3). Foreground to backgroundgeneralization also hastens rendering time – an impor-tant consideration when creating interactive environ-ments (Moore 1999). Graduated generalization can alsobe applied to the vertical axis of a dem, creating sceneswith more visible detail at higher elevations than at lowerelevations. This technique mimics the aerial perspectiveeffect, a visualization technique pioneered by EduardImhof, which accounts for the veiling effects of atmos-pheric haze. When aerial perspective is employed, high-lands, which are theoretically closer to the viewer, aredepicted with greater detail and contrast than the low-lands further away, enhancing three-dimensionality (Im-hof 1982).

Finally, greater amounts of resolution can be appliedselectively to small but otherwise important topographicfeatures on a dem, such as Puu Oo, Hawaii , mentionedearlier. Detailed objects tend to attract the viewer’s atten-

tion more readily than generalized objects (Garfield1970), a phenomenon that can be exploited beneficiallyfor 3-D landscape visualization (Figure 4).

Resolution BumpingResolution bumping is a generalization technique formanipulating Global 30 Arc Second Elevation Data Set(gtopo30) and other small-scale dems. The techniquealters digital elevation surfaces so that rugged, highmountains are more legible and look more natural ascompared to unmodified data (Figure 5).

When used for 3-D visualizations, unmodifiedgtopo30 data typically produce mountains with a choppyappearance. Vertical exaggeration, which is a graphicalnecessity when making small-scale landscape visualiza-tions, exacerbates the choppiness. Especially problematicare glaciated northern mountains comprising tightlypacked ridges and valleys – for example, the coast rangesof British Columbia, Canada, and the Alps – which oftenappear as an irregular texture rather than as recognizabletopography. Solitary high peaks with small surface areas,such as Mt. Shasta or the Matterhorn, spike upwards likethe Eiffel Tower. Used in combination, topographic de-tail, vertical exaggeration, and small-scale presentationare the enemies of legible mountain depiction.

Downsampling gtopo30 data to a sparser resolutionalleviates the problems outlined above. Generalized dataare better for depicting patterns within mountain rangesand are more tolerant of vertical exaggeration. Down-sampling elevation data, however, introduces new prob-lems that arguably are worse than the now-correctedoriginal problems. Knife-edged mountain ridges appearexcessively rounded, like the vacuum-formed plastic re-lief maps used in schools, and simplified valley bottomsare prone to misregistration with drainages.

The idea behind resolution bumping is simple: merg-ing low-resolution and high-resolution gtopo30 data ofthe same area produces hybrid data that combine the

Figure 3. Intra-DEM resolution variation, La Sal Mountains, Utah: (Left) Resolution gradually decreases fivefold from front toback within the scene. (Middle) Resolution decreases fivefold from highlands to lowlands. (Right) Combination: resolutiondecreases from front to back and from highlands to lowlands.



best characteristics and minimize the problems found inthe originals. Two copies of a gtopo30 file are used, onehigh-resolution and the other downsampled to a lowerresolution. These files can then be blended in Pho-toshop by a proportional amount that is controlled bythe user. This technique yields a new greyscale dem that,if merged in the right proportions, combines the reada-bility of the downsampled data with all the detail one ex-pects to find in mountainous terrain – without thegraphical noise. Resolution bumping in effect “bumps”or etches a suggestion of topographical detail ontogeneralized topographic surfaces (Figure 6). The resolu-tion-bumped data create an elevated base in mountain-ous regions, upon which individual mountains, withdiminished vertical scaling, project upwards (Patterson2001b).

The look of resolution-bumped gtopo30 is depend-ent on the resolution of the downsampled host data andthe percentage of blending with original high-resolution

data. The best results, I find, are achieved with drasticallydownsampled host data (10% to 20% of original resolu-tion) blended at a 40:60 ratio with the high-resolutiondata. When too little downsampling is used, or when theblending opacity is high, the resolution-bumped data donot appear to be appreciably different from unmodifieddata. At the other extreme, data that have been down-sampled by a large amount and blended lightly withhigh-resolution data yield topography that tends to lookpeculiar.

The blending modes in Photoshop behave differentlywhen merging raw elevation data, when compared to thegraphical images we are accustomed to editing. Normalblending mode works best for resolution bumping eleva-tion data. Normal blending mode etches both positive(peaks) and negative (canyons) topographic detail onthe downsampled host data below. By comparison, multi-ply mode only intensifies the darker negative values, ig-noring positive topography. The inverse is true for

Figure 4. Columbia River Gorge, Oregon, looking east. Topographic detail generated from higher-resolution DEM data directs theviewer’s attention to the gorge despite the looming presence of nearby Mt. Hood.

Figure 5. Resolution bumping: (Left) A 3-D scene of the western Alps created from 1-km GTOPO30 elevation data. The detailedterrain appears choppy when vertical exaggeration is applied. (Middle) Downsampling the data to 8-km resolution, althoughexcessively generalized, gives major landforms a more distinct appearance. (Right) A scene created from 1-km and 8-km elevationdata blended at a 4:6 ratio. This treatment etches a hint of topographic detail onto generalized major landforms.

94 tom patterson


lighten blending mode. Additional experimentationwith Photoshop’s other blending modes could produceinteresting, or even practical, results.

Resolution bumping is not a panacea for all demgeneralization situations. The technique does not workwell with large-scale elevation data and maps. Unlikegtopo30, showing rich topographic detail on large-scalemaps is usually not a problem. Rather, it is a lack of detailthat is of greater concern. The simplified small-scale to-pography in gtopo30 is much more tolerant of exoticdata manipulation, such as resolution bumping, thanlarge-scale dems – landscape visualizations created fromlarge-scale data closely mimic our real-world impressionsof the terrain. By contrast, few people have actually seenthe Alps, for example, at 1:2,000,000-scale. Even at smallscale, not all landscapes are suitable for resolutionbumping. Most non-mountain topographic features(eroded drainages, tablelands, escarpments, foothills)appear quite nicely without data manipulation, evenwhen large amounts of vertical exaggeration are used.Moreover, mountains that already have simplified forms,such as the fault-blocked Sierra Nevada in the UnitedStates, may not benefit from resolution-bumping.

One last caveat must be considered. When resolutionbumping is applied to a small-scale dem containing rug-ged mountains, the effect is also applied to adjacent low-lands, disproportionately generalizing the alreadymodest topography in these areas. Showing full lowlanddetail in conjunction with resolution-bumped topogra-phy in the mountains is sometimes the more desirablepresentation outcome. This effect can be accomplishedby rubber-stamping resolution-bumped data with a soft-edged brush to an unmodified dem only to mountainousareas in need of generalization.

Height ManipulationLightening or darkening a dem with Photoshop’s imageadjustment tools (levels, curves, brightness/contrast)raises or lowers surfaces respectively when the dem is lat-

er rendered in 3-D. This adjustment technique can beused to modify vertical exaggeration globally over an en-tire dem or, more interestingly, on selected topographicfeatures. For example, a mountain hosting a ski areacould be exaggerated in height above its surroundings(for ski area maps, bigger is better) by using the lassotool to draw a selection boundary with a feathered edgearound the mountain and lightening the area within thespace (Figure 7).

Going one step further, applying lightening and dark-ening within selections can create simple topographicfeatures. To create a volcanic cinder cone, draw a circu-lar selection with a feathered edge (again with the lassotool) and lighten the area inside – extruding the dem in3-D forms a cone-shaped hill. Drawing the circular selec-tion with a slightly irregular shape avoids excessive sym-metry and gives the cone a more realistic appearance.Last, contracting the initial circular selection by severalpixels and applying a smaller amount of darkening, de-presses the summit (see Figure 11, left image).

Glaciers can be depicted by manipulating elevationon a duplicated dem, positioned precisely below the orig-inal unaltered dem in Bryce. In Photoshop, on the bot-tom dem an imported selection boundary representingthe glacier’s extent is drawn or imported and filled withlighter pixels. Using a feathered selection boundary cre-ates a domed effect when the glacier is extruded inBryce. By increasing the bottom dem’s vertical exaggera-tion and lowering its position in Bryce, the virtual glacierwill protrude through the top dem, neatly intersectingthe valley walls (Figure 8).

Taking elevation manipulations to the extreme, theuser can apply solid black on a dem to form block dia-grams and cutaway views. Black represents the lowest ele-vation value. When applied to a selected portion of agreyscale dem it flattens and lowers the topography tobase level – the digital equivalent of a peneplain. Thebottommost elevation data can then be clipped from adem when rendered in 3-D. This technique allows select-

Figure 6. Resolution bumping technique shown in profile. Detailed (left) and generalized (middle) GTOPO30 data of the samegeographic area are blended in Photoshop. Varying the blending opacity determines the amount of topographic generalization.



Figure 7. Selective vertical exaggeration applied to a DEM of Crater Lake, Oregon.

Figure 8. (Right) Yosemite Valley ca. 20,000 B.P. Placing two DEMs (left), one on top of the other in Bryce, created the glacier byextruding the bottom DEM containing the glacier bulge through the top DEM of today’s landscape.

96 tom patterson


ed chunks of a dem to be cut away, making cross-section-al views or revealing hidden features beneath the surface(Figure 9).

Conversely, filling portions of dem with white abruptlyelevates these areas above their surroundings. On alarge-scale dem, filling small rectangular selections withwhite creates blocky shapes that, when extruded in 3-D,can pass for primitive buildings (best done on flat surfac-es to avoid sloped roofs). Text, point symbols, area pat-terns, and map linework can also be digitally embossedor recessed on topographic surfaces. This particulartechnique may be used for producing tactile physicalmodels, carved from dems by computer numerical con-trolled (cnc) routers (Faulkner 2002) – a potential boonfor the visually impaired and general audiences naturallyinclined to touch objects on display.

FilteringThe Gaussian blur filter is useful for more than general-izing dems. The filter works by filtering pixels (eleva-tions) through a mathematical “soft” lens controlled by aradius slider, which removes detail. When the radius issmall, the dem is merely smoothed; when the radius isexcessive and/or applied repeatedly, the dem flattens.Gaussian blur flattening, when applied to imported se-lection boundaries, yields benefits. For example, land/water boundaries on dems often do not match the sameboundaries on imagery or vector linework. This problemcreates unsightly misregistration near the shorelineswhen these data are later draped on dems. Editing thedem solves the problem. By importing a selection of wa-ter bodies taken from the geo-imagery or rasterized vec-tors (using the drag-and-drop technique described

previously) and applying maximum Gaussian blur, waterbody surfaces on the dem become perfectly flat at theirrespective elevations in concert with the draped imagery.The drawback to this technique is the occasional obliter-ation of distinctive topography immediately boundingwater bodies.

Gaussian blur used in moderate amounts has otheruses. Applied to a selected area on a slope, elevation-av-eraging produces terraces uncannily similar to those cre-ated by actual earth-moving equipment. This is a usefultechnique on large-scale dems to depict level areasaround buildings. Moreover, moderate Gaussian blur ap-plied to road selections removes excess height data fromelevated protuberances and adds data to bisected valleys,creating virtual road cuts and fills (Figure 10).

Photoshop contains many other graphical filters,some which have practical application for dem manipu-lation, particularly for the creation of landcover textures.For example, aa lava (a variety of rough-surfaced Hawai-

Figure 9. This portion of a cutaway view of Mammoth Cave, Kentucky, reveals some of the 563 kilometres of passages, depicted asthin yellow lines, that would otherwise be hidden beneath the surface. The park boundary crosses the cave cutaway on the right side ofthe scene, following the invisible surface topography.

Figure 10. Virtual road grading: (Left) Gaussian blur appliedto a road selection on a DEM. (Right) The modified DEMextruded in 3-D.



ian lava) can be simulated with a light application of thenoise filter. Median and/or Gaussian blur filtering ap-plied in addition to the noise filter creates a hummockyforest-like canopy. Rock texturing in high-mountain are-as, as a replacement for hand-drawn rock hachuring onsmall-scale maps (the impressionistic results would be in-appropriate at larger scales), is another potential use.Even filters that operate only with 8-bit data could beused in this endeavour – the coarseness would only ac-centuate the frost-shattered alpine environment. What-ever the type of landcover texture created by filteringdems in Photoshop, the application of these filters mustbe extremely light to produce a mere hint of a texture.Too much filtering permanently alters the character ofthe underlying elevation data. Backup copies are a ne-cessity because the technique requires experimentation.Mistakes will invariably occur.

PaintingEdits may be made to dems by painting directly on theirsurfaces (using the rubber stamp tool as a substitute forthe airbrush tool on 16-bit dems). DEM painting requiresmanual skills similar to those used in traditional illustra-tion and produces topography with a similar organic ap-pearance. Painting on dems is hampered by thedisconnect between the appearance of the 2-D greyscaledem (the surface that is painted) and 3-D model it willeventually become (see Figure 1). The problem is espe-cially acute when painting subtle tones that can be diffi-cult to see on the monochromatic surface of the dem.Nevertheless, with practice the user can learn to paintdems effectively in Photoshop to depict geological phe-nomena. For instance, to produce earthquake fissuresand glacial crevasses a user may simply draw a network ofthin dark lines with tapered ends – an unsteady hand is,for once, a benefit. Stream erosion is another application.Sinuous dark tones applied with multiple light brushstrokes, moving in an uphill direction, mimic the growthof drainage basins over time. If a progressively largerbrush size with soft edge is used, drainage basins expandin volume to simulate natural erosion (Figure 11).

Painting on a dem to depict temporal geologic events

as an image sequence or animation requires decisionsabout generalization that are more difficult than for sin-gle-image views. Nature is often much more complicatedthan can be illustrated conveniently. Consider the for-mation of Haleakala Crater on Maui, Hawaii, which, de-spite its name, is an erosional feature formed over900,000 years by coalescing amphitheatre stream valleys(Figure 12). To keep this visualization comprehensibleto readers, concurrent geologic events were ignored.This included sea level fluctuations that dramatically al-tered the shape of the island from that of today; differen-tial erosion rates on the windward and leeward shores;post-erosional volcanic events that repeatedly filled val-leys with lava; and the gradual lowering of the summit el-evation by about 600 metres. In a geologic visualization,less is often more.

Topographic SubstitutionIn describing hypothetical former and future land-scapes, geology texts often make comparisons to analo-gous present-day landscapes. This concept can beapplied to making geologic visualizations by cloning to-pography from one dem to another in a techniqueknown as topographic substitution. Providing that theuser obtains appropriate dems, topographic substitutionis easier than dem painting from scratch and, because itis based on actual dems, looks convincingly realistic. Top-ographic substitution is accomplished with the rubber

Figure 11. (Left) A greyscale DEM in Photoshop showing adendritic stream drainage painted on a flat tilted slope. (Right)The DEM extruded and rendered in 3-D.

Figure 12. DEM painting used to depict nascent stream erosion on ancestral Haleakala (left and middle), a dormant 3056-metreshield volcano that dominates present-day Maui, Hawaii (right).

98 tom patterson


stamp tool by cloning data between simultaneouslyopened 16-bit dems (Figure 13). All of Photoshop’sblending modes can be used at varying opacities.

The possibilities for mixing and matching topographyto create hybrid landscapes are nearly unlimited. For ex-ample, a depiction of the primordial landscape of NewYork City could be made by combining a strato volcano,cloned from a present-day dem of Alaska, onto a dem ofthe bayou country of Louisiana. Other applicationscould include showing the weather-worn AppalachianMountains in their former lofty glory by grafting a demof today’s Rockies on top of the Appalachians; mergingthe Kaibab and Coconino Plateaus together to in-fill theGrand Canyon; or placing Mt. Adams on top of CraterLake to depict ancient Mt. Mazama before it eruptedand imploded to form today’s caldera lake (Figures 13and 14).

Projection Plane WarpingThe projection plane of dems – the digital datum uponwhich 3-D terrain projects upwards – is flat. Using flat-world terrain models, however, can impede 3-D visualiza-tion. On low-elevation views the landscape looks mostdramatically realistic when the horizon is visible, but talltopographic features in the foreground often obscureimportant features deeper in the scene. Moreover, evenif the entire scene were visible, the oblique viewing anglerenders surface information difficult or impossible toread. Raising the viewing elevation usually solves theseproblems, but then the landscape looks too much like aconventional map, eliminating the advantages of 3-D vis-ualization.

A better solution is to emulate the view as seen froman aircraft. From high above the Earth the horizon is al-ways visible, yet when you shift your eyes downward the

Figure 13. Cloning topography between DEMs in Photoshop.

Figure 14. Crater Lake geologic evolution: (Left) Ancient Mt. Mazama created by cloning a DEM of present-day Mt. Adams, ananalogous strato volcano in the Cascade Range of Washington state, onto a DEM of present-day Crater Lake (middle) upwardsfrom the rim. (Right) A cloned DEM of today’s Aniakchak Caldera, Alaska, depicts the breached caldera wall – the fate of manycaldera lakes.



view gradually becomes less oblique and more planimet-ric. This effect can be brought to 3-D landscape visualiza-tion by adding convex curvature to the dem inPhotoshop and aligning the axis of convexity perpendic-ular to the 3-D viewing direction (Figure 15). The practi-cal result is a 3-D scene that combines the best of bothworlds: the foreground and middleground (where theimportant information resides) appear map-like, whilethe background appears realistic, complete with a hori-zon and sky (Figure 16, left). The technique borrowsfrom Heinrich Berann, the renowned traditional pano-ramist and classical painter from Austria, who died inlate 1999 (Patterson 2000). A variant of this technique,merging a small-scale dem with a sphere in Photoshop,yields a domed landscape (Figure 16, right) reminiscentof the popular over-the-horizon maps created by RichardEdes Harrison for Fortune magazine during the post–World War II years (Schulten 1998).

Whether one is merging a dem with a convex arc orsphere, a peculiarity of the technique is that vertical ex-aggeration of topography and projection-plane convexi-ty cannot be adjusted independently. Increase verticalexaggeration on a convex dem, and the projection planewill bow upward proportional to the height of the moun-tains. To ensure design options and to avoid productionbacktracking, I recommend exporting several convexdems that merge elevation and projection plane data byvarying amounts, choosing the dem that “looks right”when vertical exaggeration is applied in final 3-D pro-duction.

ConclusionUsing Photoshop to manipulate dems – a purpose forwhich the software is not specifically intended – has ena-bled the nps to enhance 3-D visualizations economically,ultimately benefiting park visitors by presenting land-

Figure 15. Merging a greyscale DEM (left) with a gradient representing an arc (middle) produces a convex DEM (right). 3-Dscenes created from convex DEMs can be viewed only from a direction perpendicular to the convexity, which in this case would be fromtop to bottom or from bottom to top.

Figure 16. (Left) Pipe Spring National Monument, Arizona. Warping the projection plane of the DEM on a convex arc that tiltstoward the viewer created the scene’s map-like foreground and receding horizon. (Right) California and the western U.S. WarpingDEM data on a sphere yields a domed topographic surface.

100 tom patterson


scapes and geo-spatial information in a more compre-hensible manner.

The Photoshop techniques used by the nps for manip-ulating dems will undoubtedly be superseded by dedicat-ed terrain applications that allow geo-data to be editedinteractively. One such promising application is Leveller(http://www.daylongraphics.com), which can importand export a wide range of dem file formats, import se-lection boundaries, apply Photoshop-like filtering, andallow painting and cloning on dem surfaces with real-time 3-D feedback. Another promising application isWorld Construction Set (wcs), a professional, albeitchallenging, 3-D landscape application that uses geo-coded data to render photo-realistic scenes completewith user-defined ecosystems and vegetation (http://www.3dnature.com). Using dems loaded into its data-base, wcs calculates varying amounts of topographic de-tail on-the-fly to depict background areas with less detailthan the foreground, enhancing 3-D depth and speed-ing rendering times. WCS also uses imported vectors,called Terraffectors,™ to etch drainages onto dem sur-faces, grade roads, and selectively modify vertical exag-geration.

In addition to emerging software applications, new va-rieties of dem data show promise for 3-D terrain presen-tation. DEMs derived from lidar (light detection andranging) data sets, for example, show subtle surface de-tails such as tree canopies, road cuts, and cliffs (Buckleyand Renslow 2002). These data, when used with restraintand in conjunction with generalized dems, have the po-tential to represent topography (including a veneer oftextured landcover) in a highly realistic fashion. On theother hand, critical data limitations still remain. The var-ious dem formats currently available are incapable ofmodelling vertical surfaces and overhanging topography.Some of the world’s unique landscapes – Arches Nation-al Park, Utah; Mammoth Cave National Park, Kentucky;and the vertical climbing walls of El Capitan, YosemiteNational Park, California, are beyond the representa-tional limits of dems. Perhaps future cartographers willhave at their disposal a “dem-Plus” format that can ade-quately depict these features.

The design and production of 3-D maps in general,and the representation of 3-D terrain in particular, is arelatively unknown field to the mainstream (2-D) carto-graphic community. The rules and conventions thathave guided cartographers for centuries largely do notapply to 3-D mapping, despite its growing popularity andthe need for design standards (Haeberling 2002). As car-tographers begin to embrace 3-D mapping using newsoftware applications and data formats, the dem manipu-lation techniques discussed here will, I hope, contributeto our understanding of 3-D terrain presentation andserve as a benchmark for further visual exploration.

SoftwareAdobe Photoshop http://www.adobe.com

BSmooth http://www.bsmooth.deCorel Bryce http://www.corel.comLeveller http://www.daylongraphics.comMacDEM http://www.treeswallow.com/macdem/index.html MicroDEM http://www.usna.edu/Users/oceano/pguth/

website/microdem.htmPBM+ plug-in http://http://www.nacis.org/cp/cp28/

resources.htmlWorld Construction Set http://www.3dnature.com

ReferencesBuckley, A. and M. Renslow. 2002. Surface Mapping, Data Han-

dling, and 3D Data Visualisations for Mountainous Terrainswith LIDAR-derived DEM. ICA 2002 Mountain CartographyWorkshop. Mt. Hood, Oregon.

Faulkner, L. 2002. Solid Terrain Models: A CommunicationsTool (keynote address). ICA 2002 Mountain CartographyWorkshop. Mt. Hood, Oregon.

Garfield, T. 1970. “The Panorama and Reliefkarte of HeinrichBerann.” Bulletin of the Society of University Cartographers 4/2:52–61.

Haeberling, C. 2002. What Affects the Representation of Topo-graphic Objects in 3D Mountain Maps? A Systematic Evalu-ation of Involved Graphic Aspects. ICA 2002 MountainCartography Workshop. Mt. Hood, Oregon.

Imhof, E. 1982. Cartographic Relief Presentation, ed. H.J. Steward.Berlin & New York: Walter de Gruyter.

Moore, K. 1999. “VRML and Java for Interactive 3D Cartogra-phy.” In Multimedia Cartography, eds. W. Cartwright, G.Gartner, and M.P. Peterson. Berlin: Springer-Verlag.

Musgrave, K. 1998. Procedural Fractal Terrains. Textures and Mode-ling: A Procedural Approach, ed. D.S. Ebert. 2nd ed. Cam-bridge, MA: Academic Press.

Patterson, T. 2000. A View from on High: Heinrich Berann’sPanoramas and Landscape Visualisation Techniques for theU.S. National Park Service. Cartographic Perspectives 36: 38–65.

———. 2001a. Retro Relief Shading with Adobe Photoshop anda Wacom Tablet. See http://www.nps.gov/carto/silvretta/retro/index.html

———. 2001b. Resolution bumping GTOPO30 in Photoshop:How to Make High Mountains More Legible. See http://www.nps.gov/carto/silvretta/bumping/bumping.html

Schulten, S. 1998. “Richard Edes Harrison and the Challenge toAmerican Cartography.” Imago Mundi 50: 174–88.

Webster, G. 2001. “Bryce Age.” 3D World 12: 48–50.

Résumé La manipulation des surfaces d’un ModèleNumérique de Terrain (MNT) à la manière d’une pâte àmodeler permet une amélioration notable de l’appar-ence et de la lisibilité de la topographie 3D des cartes. LeNational Park Service (NPS) américain utilise le pro-gramme d’édition d’images Photoshop pour manipuler lesdonnées MNT. L’exportation de données MNT retravail-lées à l’aide de graticiels (freeware) et de particiels(shareware) permet ensuite une modification ultérieurede scènes finales 3D dans Corel Bryce. Parmi les tech-



niques présentées se trouvent : la substitution topo-graphique, une méthode permettant la projection depaysages actuels dans le passé ou dans le futur ; l’exagéra-tion verticale sélective ; l’augmentation préférentielle derésolution, une technique développée spécifiquementpour améliorer la lisibilité des paysages de hautemontagne ; les effets de filtres et de peintures ; et em-pruntée aux maîtres paysagistes traditionnels, la créationde scènes 3D imitant les panoramas de Heinrich Berannou les vues sphériques de Richard Edes (déformation duplan de projection du MNT). La cartographie 3D demontagne, ainsi qu’un design parfait sont les pierres an-gulaires du programme cartographique du NPS.

Zusammenfassung Indem man Digitale Höhenmodelle(DHM) – wie beispielsweise die Modellierung von gesch-meidigem Ton – manipuliert, beeinflusst man das Er-scheinungsbild und die Lesbarkeit der 3D-Topographieauf Karten. Das U.S. National Park Service (NPS) verwen-det die bekannten Zeichnungswerkzeuge in Adobe Pho-toshop, um gerasterte DHM-Daten zu manipulieren. DasExportieren der modifizierten DHM-Daten mit Hilfe vonFreeware und Shareware-Software ermöglicht dieDarstellung von 3D-Szenen in Corel Bryce. Die zu disku-tierenden Techniken beinhalten topographischen Er-satz, eine Methode zum Nachbauen gegenwärtiger Land-schaften in die Vergangenheit oder in die Zukunft;Selektive vertikale Überhöhung; Das Erzeugen au-flösungsbezogener Unebenheiten, eine Technik, diespeziell zur Verbesserung der Lesbarkeit von Hochge-birgslandschaften entwickelt wurde; Zeichen- und Filter-effekte; und, indem man vom traditionellen Meister derLandschaftsbeschreibung das Erstellen von 3D-Szenenübernimmt, ahnt man die Panoramen von Heinrich Be-

rann, sowie die kugelförmigen, Vogelschaubilder von Ri-chard Edes Harrison nach, indem sie die Projektions-flächen verformen. Die einmaligen Herausforderungender 3D-Gebirgskartenerstellung und das fortlaufende Be-streben nach ausgezeichnetem Design – ein Eckpfeilerdes kartographischen NPS-Programms – sind die allum-fassenden Themen der NPS-Kartographie.

Resumen El tratamiento de superficies generadas conmodelos digitales de elevaciones (Digital Elevation Mod-el DEM), al igual que la utilización de modelos flexiblesde arcilla, permite mejorar la apariencia y legibilidad dela información sobre el relieve de los mapas. El U.S. Na-tional Park Service (NPS) utiliza las herramientas de ed-ición de imágenes incluidas en Adobe Photoshop paramanipular los datos raster de modelos digitales de eleva-ciones. Estos datos se exportan, con la ayuda de herrami-entas de uso gratuito, y permiten la generación deescenas 3D en Corel Bryce utilizando técnicas de “render-ing”. Las técnicas descritas incluyen la sustitucióntopográfica, un método que permite cambiar los paisajesactuales por los paisajes pasados o futuros; la exageraciónvertical selectiva; la resolución “bumping”, una técnicadesarrollada específicamente para mejorar la legibilidadde paisajes de alta montaña; los efectos de pintura y filtra-je; y, continuando la tradición de los maestros de repre-sentación del paisaje, la creación de escenas 3D queemulan los panoramas de Heinrich Berann y las vistas es-féricas sobre el horizonte de Richard Edes Harrison de-formando la proyección plana de los modelos digitalesdel terreno. Los desafíos en la representación de car-tografía 3D de montaña y la búsqueda continua de un dis-eño de calidad son la piedra angular del programacartográfico del NPS.

special issue: ica commission on mountain cartography edited … · 2015. 9. 14. · ing engine...

Documents